A database is an associated and structured data set properly stored in a storage device in a computer. A redo log stream replication technology is generally used to implement a two-node cluster hot backup of the database. An active node database sends continuously-generated redo log streams to a standby node database and the standby node database redoes the redo log streams to restore data so as to implement a hot backup. However, only limited log files are retained in the active node database. When a replication speed of the standby node database is excessively slow, redoing is excessively slow, a network is broken down, or the like, the standby node database cannot replicate a required log file from the active node database, and a stream replication mechanism stops. In this case, the standby node database needs to be reconstructed to continue to perform a stream replication.
In an existing method for reconstructing a standby node database, all data files and all log files in the active node database are replicated once to the standby node database so that a stream replication can continue after the standby node database restarts. However, when the active node database has a large data amount, this method for reconstructing a standby node database may be time-consuming and cause extremely high network resource consumption.